You are viewing the SNL 2017 Archive Website. For the latest information, see the Current Website.

Poster D36, Thursday, November 9, 6:15 – 7:30 pm, Harborview and Loch Raven Ballrooms

Predicting the negative: investigating the comprehension of negated sentences in an event-related potential study

Viviana Haase1, Markus Werning1;1Institute for Philosophy II, Ruhr University Bochum, Germany

It is widely agreed that prediction is a key feature of language comprehension (e.g. De Long et al., 2005; Pickering and Garrod, 2007) leading to the question of what we predict once we encounter a negation in a sentence. Negative sentences have been claimed to be harder to process due to (i) their higher (morpho-)syntactical complexity and (ii) the need to suppress positive information and to eventually represent the negated state of affairs (i.e. the affirmative counterpart) on a first step before representing the actual state of affairs (e.g. Just & Carpenter, 1971; Kaup et al, 2006). Accordingly, negative sentences have been shown to elicit different behavioral and neurocognitive responses than their affirmative counterparts, such as for example higher error rates, longer response times (e.g. Just & Carpenter, 1971), but also different ERP (e.g. Fischler et al, 1983; Lüdtke et al, 2008) and fMRI patterns (e.g. Bahlmann et al, 2011). Measuring ERPs, we addressed the questions (i) how prediction influences the comprehension of negated sentences, (ii) whether negation can be processed incrementally or whether a multistep process is necessary and (iii) whether processing differences between negative and affirmative sentences are correlated with individual personality or cognitive traits. Therefore, we used a 2x2 design with the factors sentence polarity (affirmative vs. negative) and truth (true vs. false). Other than Fischler et al. (1984) we did not ask for explicit truth-value judgement. Furthermore, we controlled for the lexico-semantic relationship between the two nouns occurring in a sentence, as indicated by Latent Semantic Analysis (Landauer, Foltz & Laham, 1998). Additionally, we tested our participants for their working memory capacities by employing the Digit Span and Reading Span test. Based on findings by Nieuwland et al (2010), all subjects were screened with the Autism Spectrum Quotient Questionnaire in order to examine whether differences in the results are triggered by differences in their AQ scores. We observe a clear centroposterior N400 effect for true vs false negatives, indicating that the integration of the negative marker indeed seems to be time-consuming. Other than Fischler et al. (1984) and Lüdtke et al. (2008) we argue that this result does not necessarily reflect different representational steps but rather a difficulty in the prediction process, as prediction involves computations that may require different amounts of time (e.g. Chow et al 2016). Furthermore, our results show that this N400 effect is stronger for subjects with high working memory capacities indicating that they seem to predict a higher number of possible scenarios, possibly in a graded or probabilistic way. Finally, it is argued that we should take into account language-specific differences, as different languages vary with regard to their (default) relative weighting of top-down and bottom-up information sources (e.g. MacWhinney et al 1984, Tune et al 2014). For example, preverbal and postverbal negation affect prediction to different degrees, as the latter renders a reanalysis (from an “affirmative first” reading to a sentence of negative polarity) more likely than the former.

Topic Area: Meaning: Combinatorial Semantics

Back to Poster Schedule